2. Discussing Matters Prior to the Elections

Forming pre-electoral agreements is not a mere “flavor of the month” issue in either parliamentary or presidential regimes. Pre-electoral commitments have been around at least since the end of World War II in parliamentary polities (Golder Reference Golder2005), whereas trails of pre-electoral pacts can be traced back to 1925 for their presidential counterparts (Borges et al. Reference Borges, Turgeon and Albala2021; Kellam Reference Kellam2015). Most importantly, multiparty pre-electoral coordination is still a trend in most recent elections in both parliamentary and presidential systems (Ibenskas Reference Ibenskas2016; Spoon and West Reference Spoon and West2015).

Early scholarship on pre-electoral coalitions in parliamentarism has revealed that potential governments that coalesce at an early stage of the electoral cycle are more likely to form as actual governments than purely post-electoral governments (Martin and Stevenson, Reference Martin and Stevenson2001; Strøm et al. Reference Strøm, Budge and Laver1994). The rationale is pretty consolidated: political parties engage in pre-electoral bargaining to increase their likelihood of either forming or being a part of the upcoming government (Golder Reference Golder2006; Debus Reference Debus2009; Ibenskas Reference Ibenskas2016). In other words, parties join efforts and resources once they rationalize they can form the government together. The point is that parties expect to receive more votes in general elections when they form pre-electoral alliances than when they compete on their own (Allern and Aylott Reference Allern and Aylott2009; Christiansen et al. Reference Christiansen, Nielsen and Pedersen2014). In general, in this context, parties coalesce around other parties with not-so-distant ideological preferences, and this is so for two solid reasons. First, parties strive not to lose potential voters to other parties or coalitions. Second, it is much more challenging to strike policy agreements when parties disagree over several issues than when coalition partners have preferences close to one another (Cutler et al. Reference Cutler, Scott De Marchi, Hollenbach, Laver and Orlowski2016; Golder Reference Golder2010).

The picture is quite different when we take a glimpse at presidential regimes. Initial research on presidentialism would deem the construction of pre-electoral agreements as unreasonable. The president-elect and her party would have no incentive to abandon some presidential perks in favor of their pre-electoral coalition partners since powers are fundamentally independent of one another (Stepan and Skach Reference Stepan and Skach1993), and the presidential election is basically a zero-sum game (Linz Reference Linz1990, Reference Linz, Linz and Valenzuela1994). Moreover, as presidents dispose of constitutionally fixed terms, even if they renege on their promises, the parties that comprised the pre-electoral coalition would not be able to expel them from office earlier than expected (Cheibub Reference Cheibub2007; Samuels and Shugart Reference Samuels and Shugart2010).Footnote ¹

Nonetheless, as shown previously, pre-electoral alliances are not rare in presidential democracies. In fact, Albala (Reference Albala2021) even goes so far as to say that presidential polities have far more coalition cabinets derived from pre-election alliances than their parliamentary counterparts. Therefore, how does the literature explain the emergence of pre-electoral coalitions in presidentialism? More precisely, why would a party with a competitive presidential candidate search to make pacts with other partisan organizations to back its own candidacy for the presidential office? Contrariwise, why would parties prefer to support someone else’s application for the presidency rather than launching their own candidate?

To flesh out the reasons behind the construction of pre-electoral coalitions in presidential systems, I start by addressing the first question. First, mirroring what occurs in parliamentary regimes, parties with competitive presidential candidates aim to form multiparty alliances in order to boost the vote share of their respective candidacies in the election looming on the horizon. Indeed, recent scholarship has brought to attention how presidential tickets strategically include vice-presidential candidates with specific traits to expand their potential voter base (Lopes Reference Lopes2022). This is the case as even in the absence of a viable presidential candidate, other parties might still provide politicians well-suited to a vote-seeking strategy in the presidential arena as vice presidents. This is only a single instance of how parties in a pre-electoral coalition combine different kinds of assets to leverage their odds of winning the presidential election.Footnote ²

Outside vote-seeking considerations, engaging in pre-electoral alliances also enhances the presidential party’s likelihood of securing a legislative majority in the aftermath of the election (Borges et al. Reference Borges, Turgeon and Albala2021; Carroll Reference Carroll2007). Despite minority governments not being stripped out of their governability (Strøm Reference Strøm1990), it bears noting that presidential parties have compelling incentives to look for a majority parliamentary basis even prior to the elections. Majority status confers governments with higher capabilities of passing their legislative agenda, thus circumventing possible stalemates in the legislature and making governability more straightforward (Amorim Neto et al. Reference Amorim Neto, Cox and McCubbins2003; Cheibub et al. Reference Cheibub, Przeworski and Saiegh2004; Hiroi and Renno Reference Hiroi and Renno2014; Kim Reference Kim2008). In addition to increasing the likelihood of forming majority governments, pre-electoral coalitions are also the foundation for longstanding coalition cabinets (Albala et al. Reference Albala, Borges and Couto2023). Accordingly, the formation of pre-electoral coalitions grants legislative support for a long time to presidents to get their bills approved.

Hence, political parties with competitive presidential candidates have clear-cut reasons to pursue pre-electoral agreements. Still, we have not addressed the other side of the coin. Why do parties relinquish running in the presidential elections on their own? In brief, the response lies in the fact that parties are able to reap votes, policy and office benefits from being a member of a successful pre-electoral coalition, whereas they could have gotten out of the presidential contest with empty hands had they chosen to launch a weak candidate.

The premise is that political parties without presidential aspirations do not abide by pre-electoral agreements at no cost. To start with, support in the presidential elections might come in exchange for benefits in elections at other levels, notably in gubernatorial, senatorial, and congressional electoral disputes (Borges Reference Borges2019; Borges and Turgeon Reference Borges and Turgeon2019). In this sense, some parties deliberately opt not to run for the national majoritarian election in order to focus on other electoral disputes (Borges et al. Reference Borges, Albala and Burtnik2017; Spoon and West Reference Spoon and West2015). In return, presidential parties might directly or indirectly endorse their partners’ contestants in other electoral races by withdrawing their own candidates, for instance. In fact, this was a standard procedure in the Chilean center-left coalition Concertación in the wake of the fall of Augusto Pinochet (Albala Reference Albala2013; Siavelis Reference Siavelis2002).

Additionally, in a similar vein to parliamentary coalition agreements (Moury Reference Moury2011), parties constrain the president-elect to keep her electoral policy promises (Kellam Reference Kellam2017). Although governing coalitions do not necessarily form and display written agreements in presidential systems, the enacted public policy might be close to the preferences of pre-electoral parties because presidents might feel compelled to fulfill their electoral pledges and avoid disappointing their voters.

To conclude, pre-electoral coalitions also envision distributing office rewards to their members (Carroll Reference Carroll2007). As a result, parties engage in pre-electoral agreements while knowing beforehand that they will probably have a cabinet position if the pre-electoral coalition succeeds in the presidential election.

Yet, pre-electoral agreements are not set in stone. A colorful example is that quite often pre-electoral coalitions are enlarged to accommodate other parties in post-electoral settings (Albala Reference Albala2017; Freudenreich Reference Freudenreich2016). This article points out a nuance around the government formation hitherto not explored in presidential regimes. The whole procedure of forming governments does not occur in a vacuum; instead, it takes place when parties are spread across the political spectrum. Put differently, the government formation game takes place in party systems where political parties are more or less ideologically separated from one another. This study puts forward the idea that formateur parties are more constrained to build the government around pre-electoral coalitions to the degree to which ideological polarization increases in the legislature, and the following sections explain why this should be so.

3. Legislative Polarization and Government Formation

In general, polarization is a broad concept that refers to the distance between groups regarding their stance on a specific issue. In recent years, research on polarization has focused heavily on affective polarization (e.g., Garzia et al. Reference Garzia, da Silva and Maye2023), but other topics have also attracted attention, such as activist polarization (Collitt and Highton Reference Collitt and Highton2021) and mass polarization (Levendusky Reference Levendusky2009). In this article, I am more concerned with ideological polarization in the legislature.

Legislative polarization depicts how far political parties are ideologically distant from one another in the legislature of a given party system. By and large, party systems have varying levels of legislative polarization over time. There is only a single instance where polarization is null: when all parties share the very same political preferences. However, this is hardly the case in any democratic regime. To demonstrate this point, the left panel of Figure 1 shows the average level of legislative polarization in Latin America in the period under study. In a complementary manner, to show how legislative polarization is not stationary over time, the right panel of Figure 1 illustrates the degree of ideological polarization at the party system level from 2016 to 2019 in the same region. As can be seen, Brazil’s 2018 and Chile’s 2017 general elections resulted in a degree of polarization above the countries’ respective averages, as opposed to Colombia’s 2018 and Panama’s 2019 general elections, which were below their countries’ average levels of ideological polarization.

Figure 1. Legislative Polarization in Latin America. Note: Information on the size and ideology of parties can mainly be found in the Dataset of Parties, Elections and Ideology in Latin America (DPEILA) (Borges et al. Reference Borges, Lloyd and Vommaroforthcoming). Legislative polarization has been measured by means of Dalton’s index and runs from 0 to 10 (see below for more information).

At the beginning, coalition theories ruled out the influence of ideological preferences on the government formation process. Rooted primarily in office-seeking assumptions, scholars argued that actual governments should consist of either minimal winning coalitions or coalitions with the fewest possible number of parties to retain a legislative majority (Leiserson Reference Leiserson1966; Riker Reference Riker1962). In short, either form would emerge as a consequence of parties’ unwillingness to share the spoils of government with more parties than needed. However, initial models of government formation suffered from dismaying predictive power and frequently failed to explain the process underlying the rise of governments (De Winter et al. Reference De Winter, Andeweg and Dumont2002; Laver and Schofield Reference Laver and Schofield1990). As a result, coalition theories rapidly acknowledged that political parties are also pushed by policy incentives (Axelrod Reference Axelrod1970; De Swaan Reference Swaan1973).

More recently, most studies include both office- and policy-seeking propositions in their models of coalition formation (Druckman et al. Reference Druckman, Martin and Thies2005; Eppner and Ganghof Reference Eppner and Ganghof2017; Freudenreich Reference Freudenreich2016; Giannetti and Pinto Reference Giannetti, Pinto, Albala and Reniu2018). The background is that potential cabinets marked by high ideological division are far less likely to form than alternatives that are ideologically homogeneous. Despite taking policy penchant seriously, the literature still has a tendency to resort to a crude measure of the policy-seeking approach (Indridason Reference Indridason2011). In general, the most studies on government formation operationalize ideological division as the distance between the most left-wing and the most right-wing parties within each potential government, thereby leaving aside the overall ideological differences among the parties comprising the party system.

As such, this measure entails one major problem: it disregards the general polarization of party systems. To see how this can be consequential, consider Piñera’s first cabinet in his first term in Chile. Figure 2 displays the percentage of seats in the lower chamber and the position of each political party in Chile at the time along the economic left-right dimension.Footnote ³

Figure 2. Piñera’s Government Formation Process. Note: Information on the size and ideology of parties can be found in the V-Party dataset (Lindberg et al. Reference Lindberg, Düpont, Higashijima, Kavasoglu, Marquardt, Bernhard, Döring, Hicken, Laebens, Medzihorsky, Neundorf, Reuter, Ruth-Lovell, Weghorst, Wiesehomeier, Wright, Alizada, Bederke, Gastaldi, Grahn, Hindle, Ilchenko, Von Römer, Wilson, Pemstein and Seim2022). The original 7-point scale was transformed into a 10-point scale for the sake of better visualization. There was no available information on minor and regionalist parties, such as the Independent Regionalist Party (PRI, Partido Regionalista Independiente).

On the left, Figure 2 shows the composition of both Piñera’s pre-electoral coalition and first cabinet, along with their distribution along the ideological dimension and their share of seats. As can be seen, the coalition was comprised of two right-wing parties: the National Renewal (RN, Renovación Nacional), and the Independent Democratic Union (UDI, Unión Demócrata Independiente). Likewise, the ideological position of each party present in the party system and their respective size in the legislature can be seen on the right side of Figure 2.

By and large, looking at the bigger picture of party systems provides us with further insight to make sense of government formation in presidential democracies, and this is demonstrated by the Chilean experience in 2010. From a theoretical standpoint, at the government formation stage, a pre-electoral coalition between the RN and the UDI leading to a cabinet composed of these two parties is entirely reasonable, especially considering the evolution of party competition in the country since re-democratization. However, the lack of proper attention to polarization glosses over how the party system was divided into two different camps. As a result, the RN could hardly afford to expel its pre-electoral partner to build a totally different governing coalition.

Nonetheless, not every presidential party is cloistered in such a constricted situation. To provide an example of the reverse scenario, Figure 3 illustrates the government formation process for the first cabinet of Dilma Rousseff in that same year in Brazil. In this example, we witness some pre-electoral members not being rewarded with cabinet posts in a political landscape with less pervading legislative polarization.Footnote ⁴

Figure 3. Rousseff’s Government Formation Process. Note: Information on the size of parties can be found in the DPEILA (Borges et al. Reference Borges, Lloyd and Vommaroforthcoming), and parties’ ideology hails from Baker and Greene (Reference Baker and Greene2011). The original 20-point scale was transformed into a 10-point scale for the sake of better visualization. In the left panel, the parties in dark gray did not receive a cabinet post when Rousseff assumed the presidency. In the right panel, the party in light gray was given a ministry even if it was not a member of the pre-electoral coalition.

As I shall elaborate in the next section, the bottom line is that legislative polarization and pre-electoral coalitions constrain the government formation to the point where the decision as to whom to invite to the cabinet does not depend solely on the parties that ultimately were invited to take a seat but also on the other parties available in the party system (Indridason Reference Indridason2011, 692). That is, among other things, in both instances, the formateur parties’ decision to form their cabinets was made consciously after grasping how far parties were apart in the party system and considering that there was already a pre-electoral coalition up and running.

4. The Entanglement between Pre-Electoral Coalitions, Legislative Polarization, and Government Formation

The vast majority of the literature on pre-electoral coalitions in presidential democracies argues that being part of a pre-electoral pact matters for portfolio allocation in the post-election scenario (Albala Reference Albala2021; Albala et al. Reference Albala, Borges and Couto2023; Borges et al. Reference Borges, Turgeon and Albala2021; Carroll Reference Carroll2007; Freudenreich Reference Freudenreich2016; Peron Reference Peron2018). I take a step back and claim that the degree of legislative polarization has a decisive impact on converting pre-electoral coalitions into coalition cabinets.

Legislative polarization is known for increasing the complexity around multiparty bargaining in parliamentary regimes. This is shown by the fact that governments take longer to form as legislative polarization increases (Falcó-Gimeno and Indridason Reference Falcó-Gimeno and Indridason2013; Golder Reference Golder2010; Martin and Vanberg Reference Martin and Vanberg2003). This clearly cannot happen under presidentialism because both the executive and the legislature have constitutionally fixed terms (Linz Reference Linz, Linz and Valenzuela1994), which means that governments have not only a date to end, but also a date to begin their terms. Nevertheless, this does not preclude legislative polarization from disturbing the government formation in presidentialism.

In the first place, multiparty negotiations do not represent a straightforward endeavor in fragmented party systems. Naturally, even if the presidential party still has the upper hand in many respects, sharing the power with other parties already entails bringing more players liable to veto to the scene (Tsebelis Reference Tsebelis1995). Yet, they become inherently more difficult as legislative polarization increases since parties hold increasingly irreconcilable views on various issues. Consequently, highly polarized settings present presidential parties with a smaller set of viable alternative governments, thereby reducing their leeway to build their cabinets. Conversely, slight legislative polarization represents the best scenario for the executive once it has a great variety of feasible coalition alternatives.

My point is that pre-electoral agreements counteract the effect of legislative polarization on cabinet formation. Although polarization implies more bargaining complexity, pre-electoral agreements make parties abide by several compromises even before the elections take place. In the midst of these compromises, parties discuss common ground regarding public policies to be implemented, which policies should be left aside, and which ministries should be distributed amongst coalition members (Peron Reference Peron2018). Thus, presidential parties have significant incentives to keep to their side of the bargain under polarized contexts and not dispatch any pre-electoral coalition member from the upcoming coalition government. The rationale is straightforward: building the new government around a previous, settled pre-electoral agreement is much simpler than finding the middle ground among other arrays of parties in an inhospitable party system.

Note that my contention does not imply that legislative polarization leads to greater or lesser formation of pre-electoral coalitions. On the contrary, my claim starts from the fact that pre-electoral pacts have already been made and, subsequently, constrain the government formation process increasingly more to the extent that formateur parties face greater ideological hurdles in the legislature. Also, I do not argue that parties far apart from one another cannot be part of the same pre-electoral pact. Even if their ideological positions are starkly different, they can make concessions to each other and meet halfway. This can be exemplified by the pre-electoral coalition formed between the National Convergence (CN, Convergencia Nacional) and the Movement toward Socialism (MAS, Movimiento al Socialismo) in 1993 Venezuela, two parties located at opposite ends of the ideological spectrum. In spite of the distance between the parties in the same pre-electoral pact, my argument continues: the CN had more incentives to build the government around the pre-electoral coalition to the extent that legislative polarization was moderately acute in the party system.Footnote ⁵ In the end, the CN preferred to form the government with the MAS as opposed to inviting other right-wing parties into the cabinet.

By contrast, meagre polarization along economic and social policy lines provides fewer incentives for presidents to form governments based on pre-electoral alliances. Consider the following chain of events. To start, as legislative polarization decreases, parties become less differentiated from one another and, as a consequence, have fewer disagreements over policy issues. In this sense, coalition bargaining is more likely to be undertaken and might have many different outcomes. The greater resemblance in the party system ultimately favors presidential parties since they can forgo their original pre-electoral pact and strive to build a bargain more beneficial to themselves. In a hypothetical scenario, where the party system would look like an undifferentiated amalgam of parties from similar ideological positions, presidents could dissolve their pre-electoral coalition and, rather than building a multiparty cabinet, decide to govern through ad hoc coalitions.

This discussion relates to the question of fairness in coalition governments and returns to the conundrum of whether presidents share office payoffs with pre-electoral coalition parties. Some scholars suggest that coalition governments under parliamentarism have an internalized norm by which executive office positions are proportionally allocated in relation to each party’s size in the legislature (Browne and Rice Reference Browne and Rice1979; Browne and Frendreis Reference Browne and Frendreis1980). The rationale is that proportional portfolio allocation does not derive from a purely rational approach but rather from a social norm about fairness.Footnote ⁶ This reasoning could be roughly applied to pre-electoral agreements under presidentialism, where one could argue that pre-electoral coalitions should naturally transform into post-electoral coalition cabinets. Coalition cabinets fully composed of pre-electoral coalition members should be the fairer outcome among all possible alternative governments once all parties ceasing to launch a presidential candidate would still be compensated by being granted participation in the next government.

Arguments based on norm-driven behaviors, though, remain untested in the studies on government formation in presidential regimes. The literature is, nevertheless, split into different explanations based on the rational choice theory. On the one hand, presidential parties lack incentives to maintain a bargain struck prior to the elections since their survival is not reliant upon the legislature (Kellam Reference Kellam2017). On the other hand, presidents honor the pre-electoral pact because, in rational choice terminology, governing is a repeated interaction between presidents and parties in the legislature, in which presidential parties reap the benefits of keeping their word in pre-electoral agreements by showing that they are credible coalition partners (Borges et al. Reference Borges, Turgeon and Albala2021).Footnote ⁷

These contradictory claims can be illustrated through the Brazilian case. Following the first stream of studies, as office-seeking would be out of the question, parties should join pre-electoral coalitions based solely on policy-seeking considerations (Kellam Reference Kellam2017). In this way, pre-electoral agreements should not thrive in Brazil as office-seeking parties abound in the country (Borges Reference Borges2021). However, much to the contrary, the Brazilian presidential elections have been inundated with pre-electoral coalitions since the return to democracy in 1985. More surprisingly, several party leaders only agree to engage in multiparty bargaining if they can have an eye on portfolio distribution, even if the presidential and legislative elections have not yet taken place (Peron Reference Peron2018). Taken together, the Brazilian experience has shown that parties without viable presidential candidates do expect office perks by joining a pre-electoral pact. Of course, though, anecdotal evidence from a single-case study should still be taken with a grain of salt.

That being said, my theory adds a nuance to the discussion about pre-election coalitions and their post-electoral fulfillment. Although the literature has mainly supported the idea that pre-electoral agreements matter once the government is in place, Kellam (Reference Kellam2017) still has a point when she argues that presidents might enjoy their constitutional privileges and try to exploit the office payoffs of being the formateur of the coalition. In other words, presidents might break away from pre-electoral arrangements when they can construct a more beneficial bargain for themselves. This should be most likely to happen when the legislative polarization is low, where presidential parties have more feasible alternative governments to build than they would have when parties are far apart from one another in the standard left-right dimension. This is in line with previous studies that have emphasized how presidents resort to institutional features to favor themselves in the coalition formation and lawmaking process (Inácio and Llanos Reference Inácio and Llanos2015; Silva Reference Silva and Medina2023). By contrast, presidential parties have compelling incentives to form governments around pre-electoral agreements when the ideological polarization in the legislature is high, as pre-electoral pacts serve as focal points that reduce bargaining costs, especially in comparison to forming a new government from scratch. Thus:

Hypothesis: The marginal effect of pre-electoral composition on cabinet formation grows in strength as legislative polarization gets larger; this positive effect is most substantial when legislative polarization is at its highest and vanishes as legislative polarization decreases.

5. Research Design

In order to test my argument, I analyze patterns of government formation in 13 Latin American countries. A comparative research design appears well-suited to the task as coalition governments are quite common in presidential regimes (Cheibub Reference Cheibub2007; Cheibub et al. Reference Cheibub, Przeworski and Saiegh2004), especially in Latin America (Chaisty et al. Reference Chaisty, Cheeseman and Power2018; Couto et al. Reference Couto, Soares and Livramento2021). In other words, coalitional presidentialism is not a whim of a handful of countries but rather is a real tool to engender legislative majorities in the region.

My focus resides in Latin America for a few reasons. First, focusing on a single region helps to preserve the unit homogeneity of the research (King et al. Reference King, Keohane and Verba1994). Otherwise, the results could be biased if the study had drawn on presidential regimes across different continents once non-observable features could be at play. At the same time, however, it could be said that the same logic discussed above regarding pre-electoral coalitions and legislative polarization could be extended to encompass presidential government formation outside Latin America. This leads to a second reason for centering the analysis in the Latin American context. Despite a few remarkable exceptions (Ariotti and Golder Reference Ariotti and Golder2018; Hanan Reference Hanan2012; Kim Reference Kim2011), data on presidential coalition governments in regions other than Latin America are not aggregated yet, thereby hindering comparative enterprises. Indeed, pre-election agreements have been scarcely addressed in African, Asian, and Eastern European presidential countriesFootnote ⁸ in contrast to the well-documented evidence when it comes to Latin American countries (Albala Reference Albala2021; Borges et al. Reference Borges, Turgeon and Albala2021; Freudenreich Reference Freudenreich2016; Kellam Reference Kellam2017).

To effectively test my hypothesis, I follow Freudenreich’s (Reference Freudenreich2016) lead and employ conditional logit models to study the patterns of government formation in presidentialism. Most remarkably, models based on conditional probabilities have a close-knit relationship with the literature on government formation under parliamentarism (e.g., Martin and Stevenson Reference Martin and Stevenson2001). As Freudenreich (Reference Freudenreich2016, 90) well noted, though, studies on presidential regimes have not followed the same methodological approach. A plausible reason for such a difference is that presidential systems significantly restrain the set of potential cabinets once the presidential parties are, most of the time, the formateur parties.Footnote ⁹ This institutional feature might have prompted scholars to consider only formed presidential cabinets in their analyses regarding the characteristics of presidential cabinets, such as their status in the legislature (e.g., Figueiredo et al. Reference Figueiredo, Canello and Vieira2012).

The most popular statistical techniques to deal with coalitional presidentialism, however, provide mis-specified estimates to grasp patterns of government formation. This happens because the structure of government formation makes the likelihood of forming a specific cabinet contingent on the other potential governments that could have come to power. In this way, government formation is a choice problem in which presidential parties contrast the utility of forming each alternative government with one another. To put it in presidential terminology, presidents have a whole set of possible governments from which they can choose only a single instance to come into existence. That is, just like passengers choose among different transportation systems to get to a destination (McFadden Reference McFadden1974), presidents are confronted with varying alternatives from which they have to choose one.

The dependent variable depicts whether the potential government was formed by assigning one to formed governments and zero to all others that remained only in the theoretical plan. In other words, my dependent variable is an indicator highlighting which government emerged from the coalition formation bargaining process.Footnote ¹⁰ It is worth noting that, since my claim pertains to the transition from pre-electoral coalitions to coalition cabinets, I examine only government formation opportunities that follow legislative or presidential elections. In this case, reshuffled cabinets are not present in what follows.

Data on government formation come from the ground-breaking work done by Freudenreich (Reference Freudenreich2016), which I have updated to cover more recent cases of cabinet formation.Footnote ¹¹ Following the standard procedure in parliamentary studies, only significant parties are taken into account for government formation processes. That is, parties with extremely minor legislative seats are excluded from the analysis since their size in the legislature does not influence interparty negotiationsFootnote ¹² (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tanen- baum2001; Sartori Reference Sartori1979). In practical terms, parties with less than 1% of seats in the legislature are disregarded.

Table 1 displays the countries included in the dataset, their respective temporal coverage, their number of actual governments, and their percentage of minority presidential parties.

Table 1. Dataset Summary

Information on political parties’ ideological preferences and legislative polarization comes from the Dataset of Parties, Elections and Ideology in Latin America (DPEILA) (Borges et al. Reference Borges, Lloyd and Vommaroforthcoming). The DPEILA provides the positioning of parties along the traditional economic left-right dimension by transforming the V-Party scores to a 20-point scale. In turn, ideological polarization in the legislature is measured using Dalton’s (Reference Dalton2008) Polarization Index, calculated with a slight modification from the original formula.Footnote ¹³

(1) $$Legislative\ Polarization\ Index = \sum\limits_{i = 1}^n {Pi{{\left(\;{{Si - Mi} \over {9.5}}*10\right)}^2}} $$

Where Pi is the proportion of seats of the party i, Si is the score of the party i in the left-right divide, and Mj is the mean left-right position of the party system j.Footnote ¹⁴ In plain terms, Dalton’s Polarization Index allows grasping the degree to which party systems are divided in the post-electoral scenario by weighting parties’ positions by their number of seats in the legislature.Footnote ¹⁵ More importantly, in the current ocean of different measures of polarization, Dalton’s index was explicitly built with party systems in mind. As such, it comes as no coincidence that this measurement has gained prominence among scholars in the last few years (e.g., Carroll and Kubo Reference Carroll and Kubo2021; Ecker and Meyer Reference Ecker and Meyer2015; Lupu Reference Lupu2015).

The complementing central independent variable to the analysis is Pre-Electoral Coalition (PEC), denoting whether potential governments derive their composition from pre-election coalitions. Importantly, on the one hand, potential coalitions excluding at least one of the pre-electoral coalition partners are coded as not based on pre-electoral alliances since the original multiparty pact is broken. On the other hand, alternatives containing all members or representing an extended version of pre-electoral coalitions are deemed to be based on pre-electoral multiparty bargaining.Footnote ¹⁶

Lastly, I employ most of the original variables used by Freudenreich (Reference Freudenreich2016) to serve as controls for my hypothesized claim. As a matter of fact, most of them consist of standard variables in the literature on government formation in parliamentary democracies and are broadly summarized in office, policy, and institutional incentives in the bargaining process. The first control is Minority, and it captures whether potential coalitions have majority legislative support in the lower chamber or the only chamber in the legislature.Footnote ¹⁷ Number of Parties controls for the fact that parties prefer to form governments composed of fewer as opposed to more parties by indicating how many parties are present in each potential coalition. Turning to the policy aspect of the government formation process, Ideological Division captures the distance between the most rightist and leftist party in each potential coalition, Median Party indicates whether possible governments contain the median party in their composition, and Extreme Parties does the same for extreme parties. Lastly, Runner-up Party indicates whether potential coalitions include the party that finished as the second-most voted option in the last presidential election. For more information, Table A.2 in the Supplementary Material provides more details on the operationalization of each variable in the models, and Table A.3 presents descriptive statistics.

6. Results

Table 2 provides the results for the empirical implication of my theoretical argument. The first model considers only minority government formation bargaining processes, as coalition formation is more natural when presidential parties lack a legislative majority. The second model considers all presidents, regardless of their legislative status.

Table 2. Government Formation in Latin America

Note: *p < 0.1; **p < 0.05; ***p < 0.01.

To start the analysis, as legislative polarization never reaches zero, I focus the interpretation here mainly on the interaction term (Brambor et al. Reference Brambor, Roberts Clark and Golder2007). Both models indicate that the effect of pre-electoral coalitions appears to be conditional on the degree of legislative polarization at the 0.01 level. For government formation opportunities under minority presidential parties, the first model tells us that, on average, the increase of one unit of legislative polarization makes potential governments based on pre-electoral coalitions approximately 14 times more likely to form than alternative governments that exclude any pre-election coalition partner of their composition. The same pattern holds true when majority presidents are also taken into account, as potential coalitions based on pre-election coalitions are more likely to form as legislative polarization increases.

To have a better view and a more consistent analysis of this relationship, Figure 4 plots the marginal effects of legislative polarization and pre-election coalitions on government formation across a range of values of each other based on Model 1.

Figure 4. Conditional Marginal Effect of Pre-Electoral Coalitions on Cabinet Formation.

On the left, we can see the marginal effect of legislative polarization. As the interpretation of the impact of legislative polarization on government formation requires a more comprehensive modelling strategy, as Indridason (Reference Indridason2011) does for parliamentary democracies, I refrain from exploring this effect any further. Still, this plot helps leverage information concerning the importance of pre-electoral coalitions for coalition talks.Footnote ¹⁸ In fact, by holding all else constant, the plot on the left shows that, when based on a pre-electoral coalition, potential governments are 1.3 to 3.9 times more likely to be formed than their counterparts not based on any kind of pre-electoral alliance, thereby replicating Freudenreich’s (Reference Freudenreich2016) findings when taking legislative polarization into account.

In a complementary fashion, the plot on the right provides supportive evidence for the findings in Table 2 and depicts that legislative polarization indeed conditions the effects of pre-electoral coalitions on cabinet formation. Except for low levels of ideological polarization, where the 95% confidence intervals do not let us assure the exact impact of pre-electoral pacts on the formation of presidential governments, pre-election coalitions matter most as parties are more distanced from one another on the left-right policy dimension.Footnote ¹⁹ From a Legislative Polarization Index of 1.76 onwards, the construction of pre-electoral pacts exerts increasingly more pressure on cabinet formation, increasing from 1.97 (0.30; 3.63) to 12.51 (8.38; 16.64) times the likelihood of formation of potential governments based on pre-electoral alliances. To attest to the empirical importance of this finding, 86 out of 104 formation opportunities under minority presidents analyzed by the first model score more than 1.76 in the index of Legislative Polarization, which means that disregarding legislative polarization when studying pre-electoral affairs and government formation in Latin America is quite inadvisable.

To probe whether the above findings are robust, I conduct a series of robustness tests, all of which are available in the Supplementary Material. I test whether the purported relationship between pre-electoral alliances and legislative polarization is sensitive to the distribution of upper chamber seats, the choice of ideological measure, the number of effective parliamentary parties, the occurrence of party primaries for the selection of presidential contestants, specific electoral institutions, party system institutionalization, electoral volatility, varying party–voter linkages, the use of the liberal-conservative dimension instead of the traditional economic left-right, the ideological distance between presidential and median parties, and the extent of presidential powers. Additionally, I re-run the models excluding one country each time to assess whether my results are driven by a particular country. Overall, the robustness checks yield essentially the same results as compared to those from the original models. More remarkably, the interaction between pre-electoral coalitions and legislative polarization practically never loses statistical significance and, in fact, in some models, has a more pronounced coefficient than what was previously registered. Therefore, the findings are consistent across different modelling strategies.

7. Concluding Remarks

Pre-electoral coalitions are a trademark of electoral democracies. This article has been concerned primarily with the impact of pre-electoral coalitions on government formation in presidential democracies, albeit agreements that pre-date elections are equally pivotal in parliamentary regimes (Golder Reference Golder2006; Ibenskas Reference Ibenskas2016). This focus arises from the fact that the literature on coalitional presidentialism has long been puzzled about whether pre-electoral pacts matter or not to government formation. On the one hand, the impact of pre-election coalitions as a means to distribute office benefits is belittled as formateur parties need not act in accordance with the pledges made towards other parties (Kellam Reference Kellam2017). After all, presidential parties’ survival in the executive is not reliant on the legislature (Linz Reference Linz1990, Reference Linz, Linz and Valenzuela1994; Samuels and Shugart Reference Samuels and Shugart2010). On the other hand, pre-electoral coalitions are seen as the spearheads of coalition cabinets. In this view, presidents do commit to allocating office payoffs to pre-electoral coalition members, even if they are not constitutionally obliged to stick to their pre-electoral promises (Carroll Reference Carroll2007; Freudenreich Reference Freudenreich2016; Peron Reference Peron2018).

This article has sought to bring nuance to this discussion. I argue that presidents do not behave in such a black-and-white manner. Actually, their decision to build coalition cabinets around pre-electoral agreements depends on which context the government is embedded in. More specifically, pre-electoral pacts should increasingly constrain the government formation process as legislative polarization is more pervasive in the party system. The explanation resides in the fact that an increased ideological dividedness at the party system level reduces presidents’ room to maneuver to build governing coalitions since parties have conflicting policy preferences. In this context, the utility of forming governments around pre-electoral pacts increases as they largely reduce bargaining costs. Conversely, agreements struck prior to the elections do not offer the same advantage when legislative polarization is shallow. When a party system is not composed of parties with too many disagreements on the left-right ideological dimension, presidential parties have varying possible multiparty governments at their disposal, and they may end up forming a government different from those derived from pre-electoral pacts.

To test my claim, I make use of a dataset comprising alternative governments of 13 Latin American countries over 50 years. The conditional logit models highlight, across different specifications, that the effect of pre-election coalitions is substantially conditional on the degree of legislative polarization. In other words, potential governments based on pre-electoral agreements are more likely to form as ideological polarization increases in the legislature. As a consequence, barring the formateur party, pre-electoral coalition members should not take their participation in government for granted in barely polarized settings, even if the candidate they supported won the presidential election. On the other hand, they should be more confident about being part of the government when the party system is strongly ideologically polarized in the legislature.

These findings contribute to a better understanding of government formation in multiparty presidential democracies and have particular implications for the study of pre-electoral coalitions in presidentialism. Importantly, they also reveal another way in which polarization impacts politics. While I remain neutral as to whether the moderating effect of legislative polarization on government formation is good or bad from a normative standpoint, what we should take from the results is that we should not disregard legislative polarization when studying government formation in presidential democracies.

Despite being tested on Latin American cases only, the theory put forward here applies to presidential democracies located outside the boundaries of Latin America. However, the generalizability of this work’s findings depends on how the political competition is structured in other countries. More specifically, the results of this study are meaningful for presidential democracies that have party competition at least minimally subsumed in the typical economic left-right spectrum. If this is not the case and the policy space is either multidimensional or based on non-programmatic lines, then it follows logically that we cannot extend the idea that the effect of pre-election alliances is conditional on the degree of legislative polarization.

Looking down the road, the literature would greatly benefit from taking any policy dimension other than the traditional economic left-right division into consideration, even if it means sacrificing a comparative perspective at first. For example, African and Asian presidential democracies seem to have other prominent political cleavages shaping their party systems other than the traditional left-right divide (Hanan Reference Hanan2012; Kim Reference Kim2011). Against this backdrop, the fact that not mere legislative polarization, but multidimensional legislative polarization can moderate the impact of pre-electoral coalitions in these countries merits further scholarly attention.

Additionally, despite the solid, robust results outlined above, recent deviant experiences suggest that alternative factors might also be relevant in accounting for the translation of pre-electoral coalitions into full-fledged governing coalitions. For example, even embedded in a heightened polarized scenario, the Workers’ Party (PT, Partido dos Trabalhadores) did not reward all of its pre-electoral coalition partners with ministerial portfolios when Lula was sworn into office for his third term in Brazil in 2023. Going forward, scholars would do well to expand the scope of analysis to consider features beyond the government formation process. For example, future analyses can be enriched by taking into account the alternative “tools” that governments can resort to ensure governability (Raile et al. Reference Raile, Pereira and Power2011). While only tentative speculation in nature, it can be the case that the pre-electoral coalition members deprived of cabinet posts are rewarded in another way in Brazil, such as through the favorable allocation of budgetary resources.

Likewise, there is great promise in advancing the link between pre-electoral coalitions and intra-party politics in future research. At the current stage, it is known that participation in coalition governments might trigger dissatisfaction within parties. In Venezuela, for instance, the moderate and the radical factions of the Movement for Socialism (MAS, Moviemiento al Socialismo) vehemently opposed each other’s view over participation in Caldera’s government in the mid-to-late 1990s (Handlin Reference Handlin2017). As the formation of pre-electoral coalitions implies different gains and losses for various sectors within each party, future investigations should gain from a closer examination of intra-party dynamics.

Finally, it should be stressed that pre-electoral coalitions not only influence party and electoral systems but also the accountability between voters and parties. First, as well noted by Spoon and West (Reference Spoon and West2015, 401), even if pre-electoral coalitions may be mutually beneficial to pre-electoral coalition members, they are not necessarily a blessing for representation. Much to the contrary, multiparty electoral coordination may preclude voters from casting a vote on their favorite option, as pre-electoral alliances reduce the number of available candidates on the election day. However, to the best of my knowledge, the link between pre-election coalitions and representation has been weakly explored thus far in presidential democracies, either from a theoretical or an empirical standpoint. Hence, future studies would thrive from closing this gap in our current knowledge, especially in presidential countries in Latin America, which are known for their low levels of partisanship.